Tiiu ban: Cong nghi thong tin - Tir dong hod - Cong nghe I'ii tru ISBN: 978-604-913-010-6<br />
<br />
<br />
<br />
NGHIEN ClTU B O DIEU KHIEN THONG MINH TREN CO SO TICH<br />
HOP MANG NO RON MOf VOI SUBSETHOOD VA iTNG DUNG<br />
<br />
Ha Manh Dao, Thai Quang Vinh<br />
Vien Cdng nghe thdng tin<br />
18-Hoang Qudc Viet, Ciu Giiy, Ha Ndi<br />
Email: hmdao(a),ioit.ac.vn. tqvinh@ioit.ac.vn<br />
<br />
<br />
Tom tat:<br />
Trong bdi bdo ndy chiing tdi se nghiin ciru thiit ki bd dieu khiin thdng minh<br />
tren ca sd tich hgp mgng na ron md vdi subsethoodfNFS). NFS gdm 5 lap sic<br />
dung cdc hdm thdnh viin md Gausian vd duo'c hudn luyen vdi thudt truyin nguac.<br />
Ddng thdi muc do dnh hudng cua cdc lugt md lin phdn kit ludn bdi cdc kit ndi<br />
md ciing duac dinh luang dug trin phep do subsethood tuang hd. Bd diiu khiin<br />
ndy cho phep phdt sinh tgp lugt md mgt cdch tu ddng tir du lieu hudn luyin thay vi<br />
su dung tri thirc chuyin gia. Bd diiu khiin NFS thi hiin nhiiu uu diim han so vdi<br />
edc bd diiu khiin su dung mgng na ron md thdng thudng. Cudi ciing bdi bdo<br />
ciing se di cdp din cdc vdn di dng dung cita bd diiu khiin NFS vd thuc hiin md<br />
phdng.<br />
Abstract:<br />
In this paper, we are design a intelligence controller based the integration<br />
Neuro-Fuzzy Network with mutual Subsethood (NFS). NFS includes five layers,<br />
which used gaussian membership function. NFS is used to train by the gradient<br />
descent algorithm. In this manner, NFS fully considers the contribution of input<br />
variables to the joint firing strength of fuzzy rules. Afterwards, the investigated<br />
fuzzy neural network quantifies the impacts of fuzzy rules on the consequent parts<br />
by fuzzy connections based on mutual subsethood. This controller allow<br />
automatically generate fuzzy rules from training data instead using expert<br />
knowledge. The NFS controller has many better than the conventional fuzzy<br />
neural networks. Finally, To demonstrate the capability of the NFS, simulations in<br />
control area is conducted.<br />
<br />
<br />
\. Dat van de<br />
Bd dieu khien md hien dugc img dung rdng rai trong dilu khiln, nhan dang, phan ldp<br />
mau...Van de mau chdt ciia bd dieu khien md la xay dung dugc tap luat ciia bd dilu khiln.<br />
Tap luat phai dam bao bao het cac trudng hgp ciia bai toan, dam bao sd luat la tdi thilu va sd<br />
lugng tinh toan it nhat. De thiet ke bd dilu khiln md, theo truyen thdng can phai cd tri thirc<br />
chuyen gia \'e linh vuc, nhung dieu nay trong nhilu bai toan thuc te la chua dap iing dugc vi<br />
nliieu ITnh vuc viec tim chuyen gia la khd, qua trinh thu thap tri thiic mat nhieu thdi gian va<br />
chi thich hgp vdi sd dau vao it. Trong cac bai toan phiic tap, nhieu dau vao, ngudi chuyen gia<br />
nhieu khi klidng bao duac hit cac trudng hgp thuc tl dan den bd dieu khien md ban chi muc<br />
do chinh xac va pham vi giai bai toan. D I giai quyet van de nay, cac nghien cim gan day da dl<br />
<br />
<br />
122<br />
Hoi nghi Khoa hoc k\- niim 35 ndm I 'iin Khoa hoc vd Cong nghi Viet Nam - Hd Noi 10 2010<br />
<br />
<br />
<br />
xuat nhilu thuat toan tu ddng nit ra tap luat md tir tap dir lieu %'ao/ ra ciing \'di mdt sd tri thirc<br />
biet trudc. Cac thuat toan nay chu NIU dua vao \'iec phan nhdm dii lieu dau \ao \a sir dung<br />
mang no ron \'a mang na ron md.<br />
Gan da)', dl nang eao chit lugng \'a tdi uu tap luat md. cac thuat toan phat sinh tap luat md<br />
tu ddng dua tren dii: lieu \ ao/ra da dugc cai thien bang each sir dung subsethood \'a thuat tien<br />
hoa. Vdi each sir dung phep do subsethood phai ke den cac cdng trinli[2]-[15] ciia C H .<br />
Kao[2000]. Song Hengjie at al[2009]. K. A. Rasmani. Q. Shen [2002]. Sandeep Paul. Satish<br />
Kumar[2004]. Michelle Galea, Qiang Shen[2002]...,<br />
Trong bai bao nay chiing tdi se de cap den viec xa)' dung tap luat md dua tren dii lieu \'ao/<br />
ra bang each su dung mang na ron md \'di phep do subsethood va. md phdng chung.<br />
Phan tiep theo ciia bai loan bao gdm phan 2 trinh bay \'e phep do subsethood, khao sat<br />
mang na ron md tren ca sd phep do subsethood(NFS), cap nhat tham sd \a tinh toan<br />
subsethood ciia cac lien ket md[3]; phan 3 thuc hien cai dat cac ham de khdi tao NFS va huan<br />
luyen phat sinh tap luat md sir dung mdi trudng Matlab; phan 4 thuc hien md phong . danh gia<br />
md phdng. Cudi cimg phan 5 la ket luan.<br />
2. Xay dung bo dieu khien md sir dung no ron md dua tren subsethood(NFS)<br />
2.1. Phep do subsethood<br />
Phep do subsethood cd ngudn gdc tir dinh ly P)'thagorean cd the dugc dinh nghia nhu<br />
sau[l,14,15]:<br />
Cho A. B la cac tap md thugc khdng gian U vdi ham lien thuoc PA va pe tuong img:<br />
• Phep do subsethood md S(A.B) do mirc do ma A la tap con ciia B dugc dinh nghia<br />
nhu sau:<br />
<br />
y min(/^,(z/),//„(w))<br />
<br />
M(A) Z/'N)<br />
Vdi S{A,B)&[Q,V[.<br />
Sau day chiing tdi se sir dung djuh nghia \a cac tinh chat cua subsethood dinh nghTa d tren<br />
vao de tinh lien ket md ciia bd dieu khien mang no ron md.<br />
2.2. Bg dieu khien nff ron ma sir dung subsediood(NFS)<br />
2.2.1. Mo ta NFS<br />
Trong phan nay chiing tdi se khao sat bd dieu khien mang na ron md dugc de xuat bdi S.<br />
Hengjie at al.[3]. Bd dieu khien mang na ron md su dung phep do subsethood duac the hien<br />
nhu hinh ^'e 1. Cau triic ciia bd dieu khien nay gdm 5 ldp. Ldp 1 la ldp vao. ldp 2 la ldp dieu<br />
kien, moi mit la mdt gia tri ngdn ngir ciia bien \'ao. Ldp 3 la ldp luat. Ldp 4 la ldp bieu dien<br />
phin kit luan ciia luat, mdi nut la mdt nhan ngdn ngir ciia bien ra \'a nd thuc hien giai md.<br />
Ldp 5 la ldp diu ra. Cac nut ldp 2. 3 deu su dung ham thanh vien md dang Gaussian. Diem<br />
dac biet ciia mang no ron md nay so \'di cac mang no ron md khac la nd sir dung cac lien ket<br />
md di)' dii giira ldp luat \'a ldp meiih de ket luan. Cac lien ket md nay the hien mirc dp tac<br />
ddng ciia moi phan dieu kien trong moi luat den phan ket luan nhu the nao. Mdi lien ket md<br />
ciing su dung ham thanh \'ien md dang Gaussian va xac dinh mirc do gidng nhau giira nd vdi<br />
<br />
<br />
<br />
123<br />
Tieu ban: Cong nghe thong tin - Tir dong hod - Cong nghi Vii tru ISBN: 978-604-913-010-6<br />
<br />
<br />
<br />
<br />
tap md ciia luat Rk tuong img bang each sir dung phep do subsethood md. Mang diu vao va ra<br />
ciia moi mit trong cac ldp dugc trinh bay trong bang 1.<br />
<br />
Output<br />
Laver<br />
<br />
Consequent<br />
leaver<br />
<br />
<br />
<br />
Rule<br />
Laver<br />
<br />
<br />
Antecedent<br />
Laver<br />
<br />
<br />
Input<br />
Laver<br />
<br />
<br />
<br />
<br />
Hinh I. cdu true mgng na ron md su dung subsethood(NFS)<br />
<br />
Ldp Tenldp Mang dau vao Mang dau ra<br />
<br />
1 Input layer<br />
/"'=^, >^;'*=r<br />
2 Antecedent<br />
layer<br />
/,?=-U"'-c„,)^ (2, ^J' -^^''-.,^'^'
Thêm vào BST
Báo xấu
Download
Vui lòng tải xuống để xem tài liệu đầy đủ
